欢迎来到天天文库
浏览记录
ID:39772631
大小:102.24 KB
页数:13页
时间:2019-07-11
《On the Recovery Limit of Sparse Signals Using Orthogonal Matching Pursuit》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、OntheRecoveryLimitofSparseSignalsusingOrthogonalMatchingPursuitJianWangandByonghyoShimInformationSystemLaboratorySchoolofInformationandCommunicationKoreaUniversity,Seoul,Korea136-713Email:jwang@isl.korea.ac.kr,bshim@korea.ac.krPhone:82-2-3290-4842AbstractOrthogonalmatchingpurs
2、uit(OMP)isagreedysearchalgorithmpopularlybeingusedfortherecoveryofcompressivesensedsignal.Inthispaper,weshowthatiftheisometryconstantK+1ofthesensingmatrixsatisfies3、tantialimprovementovertherecentresultofDavenportandWakin,andalsoclosesgapbetweentherecoveryboundandfundamentallimitoverwhichtheperfectrecoveryoftheOMPcannotbeguaranteed.IndexTermsCompressedsensing(CS),sparsesignal,orthogonalmatchingpursuit(OMP),restrictedisometryproperty(RIP).4、ThisworkwassupportedbyMid-careerResearcherProgramthroughNRFgrantfundedbytheMEST(No.2009-0083945)andsecondBrainKorea21project.February27,2012DRAFT1OntheRecoveryLimitofSparseSignalsusingOrthogonalMatchingPursuitI.INTRODUCTIONA.OrthogonalMatchingPursuitAsasamplingparadigmguarante5、eingthereconstructionofsparsesignalwithsamplingratesignificantlylowerthantheNyquistrate,compressivesensing(CS)hasreceivedconsiderableattentioninrecentyears[1][9].Foragivenmatrix∈Rm×n(n>m),CSrecoveryalgorithmgeneratesanestimateofK-sparsevectorx∈Rnfromasetoflinearmeasurementsy=6、x.(1)Althoughthesystemisunder-determinedandhencetheinverseproblemisingeneralill-posed,duetothepriorinformationofsignalsparsity,xcanbeperfectlyreconstructedviaproperlydesignedrecoveryalgorithm.Amongmanyrecoveryalgorithmsintheliterature,greedysearchmethodssequentiallyinvestigati7、ngthesupportofthesparsesignalhasgeneratedagreatdealofinterestforpracticalbenefit.Ineachiterationofgreedysearchalgorithms,correlationsbetweeneachcolumnofandthemodifiedmeasurements(socalledresidual)arecomparedtoidentifytheelementofthesupport.Algorithmscontainedinthiscategoryinclu8、deorthogonalmatchingpursuit(OMP)[1],regularizedorthogonalmatc
3、tantialimprovementovertherecentresultofDavenportandWakin,andalsoclosesgapbetweentherecoveryboundandfundamentallimitoverwhichtheperfectrecoveryoftheOMPcannotbeguaranteed.IndexTermsCompressedsensing(CS),sparsesignal,orthogonalmatchingpursuit(OMP),restrictedisometryproperty(RIP).
4、ThisworkwassupportedbyMid-careerResearcherProgramthroughNRFgrantfundedbytheMEST(No.2009-0083945)andsecondBrainKorea21project.February27,2012DRAFT1OntheRecoveryLimitofSparseSignalsusingOrthogonalMatchingPursuitI.INTRODUCTIONA.OrthogonalMatchingPursuitAsasamplingparadigmguarante
5、eingthereconstructionofsparsesignalwithsamplingratesignificantlylowerthantheNyquistrate,compressivesensing(CS)hasreceivedconsiderableattentioninrecentyears[1][9].Foragivenmatrix∈Rm×n(n>m),CSrecoveryalgorithmgeneratesanestimateofK-sparsevectorx∈Rnfromasetoflinearmeasurementsy=
6、x.(1)Althoughthesystemisunder-determinedandhencetheinverseproblemisingeneralill-posed,duetothepriorinformationofsignalsparsity,xcanbeperfectlyreconstructedviaproperlydesignedrecoveryalgorithm.Amongmanyrecoveryalgorithmsintheliterature,greedysearchmethodssequentiallyinvestigati
7、ngthesupportofthesparsesignalhasgeneratedagreatdealofinterestforpracticalbenefit.Ineachiterationofgreedysearchalgorithms,correlationsbetweeneachcolumnofandthemodifiedmeasurements(socalledresidual)arecomparedtoidentifytheelementofthesupport.Algorithmscontainedinthiscategoryinclu
8、deorthogonalmatchingpursuit(OMP)[1],regularizedorthogonalmatc
此文档下载收益归作者所有